Approach for Supporting Food Web Assessments with Multi-Decadal Phytoplankton Community Analyses—Case Baltic Sea

Combining the existing knowledge on links between functional characteristics of phytoplankton taxa and food web functioning with the methods from long-term data analysis, we present an approach for using phytoplankton monitoring data to draw conclusions on potential effects of phytoplankton taxonomic composition on the next trophic level. This information can be used as a part of marine food web assessments required by the Marine Strategy Framework Directive of the European Union. In this approach, both contemporary taxonomic composition and recent trends of changes are used to assess their potential consequences for food web functioning. The approach consists of four steps: (1) long-term trend analysis of class-level and total phytoplankton biomass using generalized additive models (GAMs) and calculating average biomass share of each phytoplankton class from the total phytoplankton biomass, (2) comparing the current phytoplankton community composition and its long-term changes with non-metric ordination analysis (NMDS) of genus-level biomass, (3) describing which taxa (the most accurate taxonomic level) are primarily responsible for forming the biomass and for causing the possible changes, and (4) interpretation of the phytoplankton results to assess the potential effects on the next trophic level. Within step 4, special attention is given to the following characteristic of taxa: potential suitability or quality as food for grazers, harmfulness, size, and trophy. These characteristics are selected based on existing scientific knowledge on their relevance to the higher trophic levels. In this article, we present the concept of the suggested approach and demonstrate the phytoplankton analyses with multi-decadal monitoring data from the northern Baltic Sea. We also discuss the future development of the approach toward a food web index by combining or replacing the taxonomic analyses with functional trait-based approaches.